Background
Related works
Blind image quality assessment
Semi-supervised learning
Fuzzy logic
Level | Impairment | Quality |
---|---|---|
5 | Imperceptible | Excellent |
4 | Perceptible, but not annoying | Good |
3 | Slightly annoying | Fair |
2 | Annoying | Poor |
1 | Very annoying | Bad |
Methods
Semi-supervised and fuzzy framework for blind image quality assessment
A. Natural scene statistics features
B. Quality score fuzzification
C. Semi-supervised local linear embedding
D. Quality estimation
Results and discussion
Experiments and analysis
Consistency experiments
Metric | Type | JP2K | JPEG | WN | Gblur | FF | Entire database |
---|---|---|---|---|---|---|---|
PSNR | FR | 0.8962 | 0.8596 | 0.9858 | 0.7834 | 0.8895 | 0.8240 |
SSIM | FR | 0.9367 | 0.9283 | 0.9695 | 0.8740 | 0.9428 | 0.8634 |
IFC | FR | 0.9027 | 0.9047 | 0.9581 | 0.9608 | 0.9614 | 0.9106 |
VIF | FR | 0.9615 | 0.9430 | 0.9839 | 0.9744 | 0.9618 | 0.9501 |
NSS | Blind | 0.9210 | 0.3661 | 0.8217 | 0.7007 | 0.7224 | 0.4946 |
BIQI | Blind | 0.8086 | 0.9011 | 0.9538 | 0.8293 | 0.7328 | 0.8205 |
DIIVINE | Blind | 0.9220 | 0.9210 | 0.9880 | 0.9230 | 0.8880 | 0.9170 |
BLIINDS-II | Blind | 0.9630 | 0.9793 | 0.9854 | 0.9481 | 0.9436 | 0.9232 |
SRNSS | Blind | 0.9359 | 0.9391 | 0.9404 | 0.9356 | 0.9473 | 0.9318 |
BRISQUE | Blind | 0.9229 | 0.9734 | 0.9851 | 0.9506 | 0.9093 | 0.9424 |
NIQE | Blind | 0.9370 | 0.9564 | 0.9773 | 0.9525 | 0.9128 | 0.9147 |
CBIQ | Blind | 0.912 | 0.963 | 0.959 | 0.918 | 0.885 | 0.896 |
S2F2-I | Blind |
0.9578
|
0.9489
|
0.9668
|
0.9556
|
0.9370
|
0.9464
|
The performance of S2F2-I
The performance of S2F2-II
Metric | LCC |
---|---|
S2F2 5 scale | 0.9464 |
S2F2-II 5-5 | 0.9560 |
S2F2 3 scale | 0.9453 |
S2F2-II 3-3 |
0.9602
|
Robustness to the size of the training set
Robustness to parameters
Membership function
Fuzzy scales
Variance
Neighborhood selection
Robustness to databases
JP2K | JPEG | WN | Gblur | All | |
---|---|---|---|---|---|
PSNR | 0.8250 | 0.8760 | 0.9230 | 0.9342 | 0.8700 |
SSIM | 0.9603 | 0.9354 | 0.8168 | 0.9544 | 0.9016 |
BLIINDS-II | 0.9157 | 0.901 | 0.6600 | 0.8500 | 0.8442 |
DIIVINE | 0.924 | 0.966 | 0.851 | 0.862 | 0.889 |
BRISQUE | 0.832 | 0.924 | 0.82 | 0.881 | 0.896 |
S2F2-I | 0.9115 | 0.9143 | 0.7503 | 0.8422 | 0.8761 |